The one-size-fits-all policy is a problem

The data you present is somewhat compelling. An answer with many cached versions, showing the answer being fleshed out over time, is a pretty reasonable defense against ChatGPT. I imagine you will eventually turn this heuristic into a metric that mods have access to. I'm not sure I completely buy all your conclusions, but after a quick skim, the data is more compelling than I thought it would be. Certainly none of us mods want to suspend people who don't deserve it.

But different sites are different. On my site (Academia.SE), many of our askers are at a very vulnerable stage in their lives, and they're looking for career guidance from people who have been there before. This is not a programming site where they can say "thanks, but I tried it and it didn't work"; they are often facing life-changing decisions that can only be made once. It is unacceptable that they would make such decisions (unknowingly) based on auto-generated answers.

  In other words: the risk ratio on our site is completely different than StackOverflow. On StackOverflow, you are very worried about alienating potential contributors with a false accusation. But on Academia, I would much rather turn away someone who is posting content indistinguishable from ChatGPT (which we don't really want anyway) than about letting this nonsense stand. Sure, on the high-traffic posts, the nonsense gets downvoted -- but on the low-traffic, more "niche" posts, votes are much less reliable.

Our community can figure this out. We don't need you to tell us what the answer is. We need you to tell us what the problem is (which you have now done) and then we'll come up with the answer. Our goals are aligned: we don't want to suspend innocent users, and I'm sure you don't want bad advice to run rampant. So please retract your policy (at least for the smaller sites) and let us mods get back to work: we'll start a conversation with our community on meta and decide how we can better balance these two priorities.

The one-size-fits-all policy is a problem

The data you present is somewhat compelling. An answer with many cached versions, showing the answer being fleshed out over time, is a pretty reasonable defense against ChatGPT. I imagine you will eventually turn this heuristic into a metric that mods have access to. I'm not sure I completely buy all your conclusions, but after a quick skim, the data is more compelling than I thought it would be. Certainly none of us mods want to suspend people who don't deserve it.

But different sites are different. On my site (Academia.SE), many of our askers are at a very vulnerable stage in their lives, and they're looking for career guidance from people who have been there before. This is not a programming site where they can say "thanks, but I tried it and it didn't work"; they are often facing life-changing decisions that can only be made once. It is unacceptable that they would make such decisions (unknowingly) based on auto-generated answers. In other words: on Academia.SE, I would much rather turn away someone who is posting content indistinguishable from ChatGPT (which we don't really want anyway) than let someone base a difficult decision on some auto-generated nonsense.

Our community can figure this out. We don't need you to tell us what the answer is. We need you to tell us what the problem is (which you have now done) and then we'll come up with the answer. Our goals are aligned: we don't want to suspend innocent users, and I'm sure you don't want bad advice to run rampant. So please retract your policy (at least for the smaller sites) and let us mods get back to work: we'll start a conversation with our community on meta and decide how we can better balance these two priorities.